Method, apparatus and system for biometric identification

Abstract

Method and apparatus for processing a biometric measurement signal using a computing device, including receiving a biometric measurement signal generated by contact with a single individual, extracting at least one periodic fragment from the biometric measurement signal, generating first feature data at least partially based on the at least one extracted periodic fragment, determining second feature data from the first feature data by removing data from the first feature data using robust principal component analysis, determining whether a match exists between the second feature data and defined biometric data associated with a known individual by processing the second feature data and the defined biometric data using a machine learning technique, and in response to determining a match exists between the second feature data and the defined biometric data, transmitting a signal indicating that the single individual is the known individual.

What is claimed is: 
     
       1. A method for processing a biometric measurement signal using a computing device, comprising:
 receiving, by the computing device from a biometric sensor, a biometric measurement signal generated by contact with a single individual, wherein the biometric measurement signal comprises an electrocardiogram (ECG) signal; 
 extracting periodic fragments from the biometric measurement signal, wherein each periodic fragment is indicative of a complete heartbeat of the single individual and is represented by a vector, and wherein each periodic fragment is represented by an image; 
 generating, by the computing device, first feature data by aggregating the periodic fragments, wherein the first feature data comprises an aggregated matrix generated by placing vectors representing the periodic fragments side by side; 
 determining, by the computing device, second feature data by removing noise data and outliers from the first feature data using robust principal component analysis, wherein the second feature data comprises a valid data matrix generated by removing the noise data and the outliers from the aggregated matrix; 
 generating an aggregated image by stacking images of the periodic fragments associated with the second feature data; 
 determining, at the computing device, a difference image as a difference between the aggregated image and a second aggregated image representing defined biometric data; 
 inputting the difference image into a neural network, wherein the neural network is trained using difference images and associated ground truth labels, the ground truth labels representing respective relationships between two individuals; and 
 in response to the neural network outputting a label indicating that a match exists between the second feature data and the defined biometric data, transmitting, using a communication channel to a device for identification and authentication of the single individual, a signal indicating that the single individual is a known individual. 
 
     
     
       2. The method of  claim 1 , wherein the neural network comprises a convolutional neural network (CNN) model, and the second feature data and the defined biometric data are used as inputs to the CNN model to determine whether the match exists. 
     
     
       3. The method of  claim 1 , further comprising:
 receiving third feature data generated by aggregating periodic fragments extracted during a setup process from a second biometric measurement signal associated with the known individual; and 
 determining the defined biometric data by removing noise data and outliers from the third feature data using the robust principal component analysis. 
 
     
     
       4. The method of  claim 1 , wherein the biometric measurement signal further comprises at least one of an electroencephalography (EEG) signal, a photoplethysmography (PPG) signal, an electromyography (EMG) signal. 
     
     
       5. The method of  claim 1 , wherein the periodic fragments comprise PQRST fragments. 
     
     
       6. The method of  claim 1 , wherein the device for identification and authentication comprises one of the computing device and a remote server. 
     
     
       7. The method of  claim 1 , further comprising:
 ordering the valid data matrix into a row echelon form; and 
 selecting top K rows of the row echelon form as the second feature data, wherein K is an integer not greater than a rank of the aggregated matrix. 
 
     
     
       8. The method of  claim 7 , wherein the device for identification and authentication comprises a gate control. 
     
     
       9. An apparatus for processing a biometric measurement signal, comprising:
 a non-transitory memory; and 
 a processor configured to execute instructions stored in the non-transitory memory to: 
 receive, from a biometric sensor coupled to a wearable device, a biometric measurement signal generated by contact with a single individual, wherein the biometric measurement signal comprises an electrocardiogram (ECG) signal; 
 pre-process the biometric measurement signal to remove baseline wander from the biometric measurement signal; 
 extract periodic fragments from the biometric measurement signal, wherein each periodic fragment is indicative of a complete heartbeat of the single individual and is represented by a vector; 
 generate first feature data by aggregating the periodic fragments, wherein the first feature data comprises an aggregated matrix generated by placing vectors representing the periodic fragments side by side; 
 determine second feature data by removing noise data and outliers from the first feature data using robust principal component analysis, wherein the second feature data comprises a valid data matrix generated by removing the noise data and the outliers from the aggregated matrix; 
 determine a difference matrix as a difference between the valid data matrix and a matrix representing defined biometric data; 
 input the difference matrix into a neural network, wherein the neural network is trained using difference matrices and associated ground truth labels, the ground truth labels representing respective relationships between two individuals; and 
 in response to the neural network outputting a label indicating that a match exists between the second feature data and the defined biometric data, transmit, using a communication device to a device for identification and authentication of the single individual, a signal indicating that the single individual is a known individual. 
 
     
     
       10. The apparatus of  claim 9 , further comprising:
 a body of the apparatus; wherein 
 the biometric sensor coupled to the body to produce the biometric measurement signal when activated by contact with the single individual; and 
 the communication device coupled to the body and controlled by the processor to transmit the signal, to a reader device, indicating that the single individual is the known individual. 
 
     
     
       11. The apparatus of  claim 9 , wherein the instructions further comprise instructions to:
 receive third feature data generated by aggregating periodic fragments extracted during a setup process from a biometric measurement signals associated with the known individual; 
 determine the defined biometric data by removing noise data and outliers from the third feature data using the robust principal component analysis; 
 order the valid data matrix into a row echelon form; and 
 select top K rows of the row echelon form as the second feature data, wherein K is an integer not greater than a rank of the aggregated matrix. 
 
     
     
       12. The apparatus of  claim 9 , wherein the biometric measurement signal comprises at least one of an electroencephalography (EEG) signal, a photoplethysmography (PPG) signal, an electromyography (EMG) signal. 
     
     
       13. The apparatus of  claim 9 , wherein the periodic fragments comprise PQRST fragments. 
     
     
       14. The apparatus of  claim 9 , wherein the neural network comprises a deep learning (DL) model, the DL model comprises a convolutional neural network (CNN) model. 
     
     
       15. A system, comprising:
 a wearable apparatus comprising:
 a body of the wearable apparatus, coupled to a securing mechanism for securing the wearable apparatus to a portion of a single individual; 
 a biometric sensor coupled to the body of the wearable apparatus and positioned on the portion of the single individual, configured to receive a biometric measurement signal comprising an electrocardiogram (ECG) signal when the biometric sensor is activated by contact with the single individual; and 
 a first communication device coupled to the body of the wearable apparatus; 
 
 a computing device comprising:
 a second communication device, configured to communicate with the first communication device; 
 a memory; and 
 a processor coupled to the memory configured to execute instructions stored in the memory to:
 wirelessly receive the biometric measurement signal from the first communication device; 
 pre-process the biometric measurement signal to remove baseline wander from the biometric measurement signal; 
 extract periodic fragments from the biometric measurement signal, wherein each periodic fragment is indicative of a complete heartbeat of the single individual and is represented by a vector; 
 generate first feature data by aggregating periodic fragments, wherein the first feature data comprises an aggregated matrix generated by placing vectors representing the periodic fragments side by side;
 determine second feature data by removing noise data and outliers from the first feature data using robust principal component analysis, wherein the second feature data comprises a valid data matrix generated by removing the noise data and the outliers from the aggregated matrix; 
 determine a difference matrix as a difference between the valid data matrix and a matrix representing defined biometric data; 
 input the difference matrix into a convolutional neural network (CNN) model, wherein the CNN is trained using difference matrices and associated ground truth labels, the ground truth labels representing respective relationships between two individuals; and 
 
 in response to the CNN model outputting a label indicating that a match exists between the second feature data and the defined biometric data, wirelessly transmit a signal to an external server for identification and authentication of the single individual indicating that the single individual is a known individual.